Project R-11556

Event Data Analytics: Feature Engineering and Diagnostic Analysis in a Hybrid Data Source Setting (Research)

In these modern times, businesses are being confronted more and more with a large volume of fine-grained data. However, they want to extract knowledge and meaning from this data to support their business and enable data-driven growth. Event data analytics can assist by, among others, learning how to stimulate certain events to happen, while preventing others from occurring. This PhD project aims at delivering these insights. To this end, three focal point can be distinguished of which the first two constitute potential data enrichment stages. Firstly, the masses of fine-grained data must be aggregated to a more understandable level. This means transforming events to behavioural patterns by means of, for example, event log abstraction techniques. Secondly, context changes must be detected. This includes discovering relations between different business objects and their attributes. These objects include, but are not limited to, customers, suppliers, employees, IT systems, assets, … One object can influence the other, both in a positive and negative way. Thirdly, diagnostic analytics will help understand which patterns are closely related to the phenomena of interest. In other words, while diagnostic analytics typically searches for the cause of a problem so that the problem can be avoided, this work also attempts to uncover how one can stimulate a new occurrence of positive events. This can be achieved with the same techniques.

01 September 2019 - 11 January 2024